arXiv subjects

Zixin Yin

Publications and source records attributed to Zixin Yin.

17 recordsLinked to original sources

arXiv2026

Training Skills Like Parameters via Self-Supervised Semantic Diffusion

While Large Language Models (LLMs) demonstrate remarkable general instruction-following capabilities, they often fall short of human experts in highly specialized, open-ended domains such as creative screenwriting. Prior approaches typically adopt post-training, yet both supervised fine-tuning and reinforcement learning require weight access that closed-source frontier models do not offer, and demand heavy compute. Moreover, what is learned is tied to a single checkpoint and cannot be inspected by humans. Recent advancements in agentic continual learning instead attempt to bridge this gap by accumulating external textual skills. However, these methods heavily rely on costly human expert annotations or unreliable LLM-as-a-judge feedback for reflection. To overcome this bottleneck, we propose a novel, unsupervised self-evolving agent framework inspired by the corruption-and-reconstruction paradigm of diffusion models. Instead of relying on explicit external scoring, we leverage existing high-quality human artifacts to construct self-supervised signals. Training then follows the familiar loop of neural network training, forward, loss, and backward, with the loss coming from contrasting the agent's reconstruction against the human original. What is updated is not model weights but an external library of textual skills. We evaluate our framework on the challenging task of short drama screenwriting. Experimental results demonstrate that our method enables the agent to autonomously extract and internalize highly generalizable skills, significantly enhancing its domain-specific generation capabilities. Furthermore, this self-contrastive reflection paradigm offers a scalable pathway for agents to teach themselves the production of complex, high-quality human artifacts, without requiring external supervision.

cs.CL↗

arXiv2026

Motion4Motion: Motion Transfer Across Subjects at Inference

This work explores the motion transfer from one video to another, which is crucial in animation for diverse characters. Previously, video motion transfer has been largely explored between human and human-like characters, enabling a lot of applications in digital creation. However, these approaches encounter a main limitation. Specifically, related technical pipelines heavily rely on a predefined human skeleton structure and accordingly require skeleton-conditional model training. On the one hand, these methods are difficult to generalize to diverse characters, such as animals from different species, while preserving their unique motion styles. On the other hand, labeled data in diverse skeletons is limited, which additionally restricts the large-scale training for the task. In this paper, we jump out of the skeleton-based motion transfer framework and propose a training-free motion transfer framework, named Motion4Motion. Motion4Motionmodels the motion flow of the character in a video instead of skeletons, which makes motion transfer across species easier. Extensive experimental results and novel applications show our methods outperform baselines impressively. Project page is available at https://lhchen.top/Motion4Motion.

cs.CV↗

arXiv2026

SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture

Recent large vision-language models (VLMs) remain fundamentally constrained by a persistent dichotomy: understanding and generation are treated as distinct problems, leading to fragmented architectures, cascaded pipelines, and misaligned representation spaces. We argue that this divide is not merely an engineering artifact, but a structural limitation that hinders the emergence of native multimodal intelligence. Hence, we introduce SenseNova-U1, a native unified multimodal paradigm built upon NEO-unify, in which understanding and generation evolve as synergistic views of a single underlying process. We launch two native unified variants, SenseNova-U1-8B-MoT and SenseNova-U1-A3B-MoT, built on dense (8B) and mixture-of-experts (30B-A3B) understanding baselines, respectively. Designed from first principles, they rival top-tier understanding-only VLMs across text understanding, vision-language perception, knowledge reasoning, agentic decision-making, and spatial intelligence. Meanwhile, they deliver strong semantic consistency and visual fidelity, excelling in conventional or knowledge-intensive any-to-image (X2I) synthesis, complex text-rich infographic generation, and interleaved vision-language generation, with or without think patterns. Beyond performance, we show detailed model design, data preprocessing, pre-/post-training, and inference strategies to support community research. Last but not least, preliminary evidence demonstrates that our models extend beyond perception and generation, performing strongly in vision-language-action (VLA) and world model (WM) scenarios. This points toward a broader roadmap where models do not translate between modalities, but think and act across them in a native manner. Multimodal AI is no longer about connecting separate systems, but about building a unified one and trusting the necessary capabilities to emerge from within.

cs.CV↗

arXiv2025

ReasonEdit: Towards Reasoning-Enhanced Image Editing Models

Recent advances in image editing models have shown remarkable progress. A common architectural design couples a multimodal large language model (MLLM) encoder with a diffusion decoder, as seen in systems such as Step1X-Edit and Qwen-Image-Edit, where the MLLM encodes both the reference image and the instruction but remains frozen during training. In this work, we demonstrate that unlocking the reasoning capabilities of MLLM can further push the boundaries of editing models. Specifically, we explore two reasoning mechanisms, thinking and reflection, which enhance instruction understanding and editing accuracy. Based on that, our proposed framework enables image editing in a thinking-editing-reflection loop: the thinking mechanism leverages the world knowledge of MLLM to interpret abstract instructions, while the reflection reviews editing results, automatically corrects unintended manipulations, and identifies the stopping round. Extensive experiments demonstrate that our reasoning approach achieves significant performance gains, with improvements of ImgEdit (+4.3%), GEdit (+4.7%), and Kris (+8.2%) when initializing our DiT from the Step1X-Edit (ReasonEdit-S), and also outperforms previous open-source methods on both GEdit and Kris when integrated with Qwen-Image-Edit (ReasonEdit-Q).

cs.CV↗

arXiv2025

ConsistEdit: Highly Consistent and Precise Training-free Visual Editing

Recent advances in training-free attention control methods have enabled flexible and efficient text-guided editing capabilities for existing generation models. However, current approaches struggle to simultaneously deliver strong editing strength while preserving consistency with the source. This limitation becomes particularly critical in multi-round and video editing, where visual errors can accumulate over time. Moreover, most existing methods enforce global consistency, which limits their ability to modify individual attributes such as texture while preserving others, thereby hindering fine-grained editing. Recently, the architectural shift from U-Net to MM-DiT has brought significant improvements in generative performance and introduced a novel mechanism for integrating text and vision modalities. These advancements pave the way for overcoming challenges that previous methods failed to resolve. Through an in-depth analysis of MM-DiT, we identify three key insights into its attention mechanisms. Building on these, we propose ConsistEdit, a novel attention control method specifically tailored for MM-DiT. ConsistEdit incorporates vision-only attention control, mask-guided pre-attention fusion, and differentiated manipulation of the query, key, and value tokens to produce consistent, prompt-aligned edits. Extensive experiments demonstrate that ConsistEdit achieves state-of-the-art performance across a wide range of image and video editing tasks, including both structure-consistent and structure-inconsistent scenarios. Unlike prior methods, it is the first approach to perform editing across all inference steps and attention layers without handcraft, significantly enhancing reliability and consistency, which enables robust multi-round and multi-region editing. Furthermore, it supports progressive adjustment of structural consistency, enabling finer control.

cs.CV↗

arXiv2025

LazyDrag: Enabling Stable Drag-Based Editing on Multi-Modal Diffusion Transformers via Explicit Correspondence

The reliance on implicit point matching via attention has become a core bottleneck in drag-based editing, resulting in a fundamental compromise on weakened inversion strength and costly test-time optimization (TTO). This compromise severely limits the generative capabilities of diffusion models, suppressing high-fidelity inpainting and text-guided creation. In this paper, we introduce LazyDrag, the first drag-based image editing method for Multi-Modal Diffusion Transformers, which directly eliminates the reliance on implicit point matching. In concrete terms, our method generates an explicit correspondence map from user drag inputs as a reliable reference to boost the attention control. This reliable reference opens the potential for a stable full-strength inversion process, which is the first in the drag-based editing task. It obviates the necessity for TTO and unlocks the generative capability of models. Therefore, LazyDrag naturally unifies precise geometric control with text guidance, enabling complex edits that were previously out of reach: opening the mouth of a dog and inpainting its interior, generating new objects like a ``tennis ball'', or for ambiguous drags, making context-aware changes like moving a hand into a pocket. Additionally, LazyDrag supports multi-round workflows with simultaneous move and scale operations. Evaluated on the DragBench, our method outperforms baselines in drag accuracy and perceptual quality, as validated by VIEScore and human evaluation. LazyDrag not only establishes new state-of-the-art performance, but also paves a new way to editing paradigms.

cs.CV↗

arXiv2025

UniVerse-1: Unified Audio-Video Generation via Stitching of Experts

We introduce UniVerse-1, a unified, Veo-3-like model capable of simultaneously generating coordinated audio and video. To enhance training efficiency, we bypass training from scratch and instead employ a stitching of experts (SoE) technique. This approach deeply fuses the corresponding blocks of pre-trained video and music generation experts models, thereby fully leveraging their foundational capabilities. To ensure accurate annotations and temporal alignment for both ambient sounds and speech with video content, we developed an online annotation pipeline that processes the required training data and generates labels during training process. This strategy circumvents the performance degradation often caused by misalignment text-based annotations. Through the synergy of these techniques, our model, after being finetuned on approximately 7,600 hours of audio-video data, produces results with well-coordinated audio-visuals for ambient sounds generation and strong alignment for speech generation. To systematically evaluate our proposed method, we introduce Verse-Bench, a new benchmark dataset. In an effort to advance research in audio-video generation and to close the performance gap with state-of-the-art models such as Veo3, we make our model and code publicly available. We hope this contribution will benefit the broader research community. Project page: https://dorniwang.github.io/UniVerse-1/.

cs.CV↗

arXiv2025

Motion2Motion: Cross-topology Motion Transfer with Sparse Correspondence

This work studies the challenge of transfer animations between characters whose skeletal topologies differ substantially. While many techniques have advanced retargeting techniques in decades, transfer motions across diverse topologies remains less-explored. The primary obstacle lies in the inherent topological inconsistency between source and target skeletons, which restricts the establishment of straightforward one-to-one bone correspondences. Besides, the current lack of large-scale paired motion datasets spanning different topological structures severely constrains the development of data-driven approaches. To address these limitations, we introduce Motion2Motion, a novel, training-free framework. Simply yet effectively, Motion2Motion works with only one or a few example motions on the target skeleton, by accessing a sparse set of bone correspondences between the source and target skeletons. Through comprehensive qualitative and quantitative evaluations, we demonstrate that Motion2Motion achieves efficient and reliable performance in both similar-skeleton and cross-species skeleton transfer scenarios. The practical utility of our approach is further evidenced by its successful integration in downstream applications and user interfaces, highlighting its potential for industrial applications. Code and data are available at https://lhchen.top/Motion2Motion.

cs.CV↗

arXiv2025

Training-Free Text-Guided Color Editing with Multi-Modal Diffusion Transformer

Text-guided color editing in images and videos is a fundamental yet unsolved problem, requiring fine-grained manipulation of color attributes, including albedo, light source color, and ambient lighting, while preserving physical consistency in geometry, material properties, and light-matter interactions. Existing training-free methods offer broad applicability across editing tasks but struggle with precise color control and often introduce visual inconsistency in both edited and non-edited regions. In this work, we present ColorCtrl, a training-free color editing method that leverages the attention mechanisms of modern Multi-Modal Diffusion Transformers (MM-DiT). By disentangling structure and color through targeted manipulation of attention maps and value tokens, our method enables accurate and consistent color editing, along with word-level control of attribute intensity. Our method modifies only the intended regions specified by the prompt, leaving unrelated areas untouched. Extensive experiments on both SD3 and FLUX.1-dev demonstrate that ColorCtrl outperforms existing training-free approaches and achieves state-of-the-art performances in both edit quality and consistency. Furthermore, our method surpasses strong commercial models such as FLUX.1 Kontext Max and GPT-4o Image Generation in terms of consistency. When extended to video models like CogVideoX, our approach exhibits greater advantages, particularly in maintaining temporal coherence and editing stability. Finally, our method also generalizes to instruction-based editing diffusion models such as Step1X-Edit and FLUX.1 Kontext dev, further demonstrating its versatility.

cs.GR↗

arXiv2025

HH-Codec: High Compression High-fidelity Discrete Neural Codec for Spoken Language Modeling

Discrete speech tokenization is a fundamental component in speech codecs. However, in large-scale speech-to-speech systems, the complexity of parallel streams from multiple quantizers and the computational cost of high-time-dimensional codecs pose significant challenges. In this paper, we introduce HH-Codec, a neural codec that achieves extreme compression at 24 tokens per second for 24 kHz audio while relying on single-quantizer inference. Our approach involves a carefully designed Vector Quantization space for Spoken Language Modeling, optimizing compression efficiency while minimizing information loss. Building on this, we propose an asymmetric encoder-decoder architecture (Audio-VQ-Mel-Audio) that leverages dual supervision and progressive training to enhance reconstruction stability and fidelity. HH-Codec achieves state-of-the-art performance in speech reconstruction with an ultra-low bandwidth of 0.3 kbps. We further evaluate its effectiveness in codebook utilization and generative model adaptation, with extensive ablations validating the necessity of each module. HH-Codec is available at https://github.com/opendilab/HH-Codec.

cs.SD↗

arXiv2025

SpeakerVid-5M: A Large-Scale High-Quality Dataset for Audio-Visual Dyadic Interactive Human Generation

The rapid development of large-scale models has catalyzed significant breakthroughs in the digital human domain. These advanced methodologies offer high-fidelity solutions for avatar driving and rendering, leading academia to focus on the next major challenge: audio-visual dyadic interactive virtual human. To facilitate research in this emerging area, we present SpeakerVid-5M dataset, the first large-scale, high-quality dataset designed for audio-visual dyadic interactive virtual human generation. Totaling over 8,743 hours, SpeakerVid-5M contains more than 5.2 million video clips of human portraits. It covers diverse scales and interaction types, including monadic talking, listening, and dyadic conversations. Crucially, the dataset is structured along two key dimensions: interaction type and data quality. First, it is categorized into four types (dialogue branch, single branch, listening branch and multi-turn branch) based on the interaction scenario. Second, it is stratified into a large-scale pre-training subset and a curated, high-quality subset for Supervised Fine-Tuning (SFT). This dual structure accommodates a wide array of 2D virtual human tasks. In addition, we provide an autoregressive (AR)-based video chat baseline trained on this data, accompanied by a dedicated set of metrics and test data to serve as a benchmark VidChatBench for future work. Both the dataset and the corresponding data processing code will be publicly released. Project page: https://dorniwang.github.io/SpeakerVid-5M/

cs.CV↗

arXiv2022

Talking Head Generation with Probabilistic Audio-to-Visual Diffusion Priors

In this paper, we introduce a simple and novel framework for one-shot audio-driven talking head generation. Unlike prior works that require additional driving sources for controlled synthesis in a deterministic manner, we instead probabilistically sample all the holistic lip-irrelevant facial motions (i.e. pose, expression, blink, gaze, etc.) to semantically match the input audio while still maintaining both the photo-realism of audio-lip synchronization and the overall naturalness. This is achieved by our newly proposed audio-to-visual diffusion prior trained on top of the mapping between audio and disentangled non-lip facial representations. Thanks to the probabilistic nature of the diffusion prior, one big advantage of our framework is it can synthesize diverse facial motion sequences given the same audio clip, which is quite user-friendly for many real applications. Through comprehensive evaluations on public benchmarks, we conclude that (1) our diffusion prior outperforms auto-regressive prior significantly on almost all the concerned metrics; (2) our overall system is competitive with prior works in terms of audio-lip synchronization but can effectively sample rich and natural-looking lip-irrelevant facial motions while still semantically harmonized with the audio input.

cs.GR↗

arXiv2022

Progressive Disentangled Representation Learning for Fine-Grained Controllable Talking Head Synthesis

We present a novel one-shot talking head synthesis method that achieves disentangled and fine-grained control over lip motion, eye gaze&blink, head pose, and emotional expression. We represent different motions via disentangled latent representations and leverage an image generator to synthesize talking heads from them. To effectively disentangle each motion factor, we propose a progressive disentangled representation learning strategy by separating the factors in a coarse-to-fine manner, where we first extract unified motion feature from the driving signal, and then isolate each fine-grained motion from the unified feature. We introduce motion-specific contrastive learning and regressing for non-emotional motions, and feature-level decorrelation and self-reconstruction for emotional expression, to fully utilize the inherent properties of each motion factor in unstructured video data to achieve disentanglement. Experiments show that our method provides high quality speech&lip-motion synchronization along with precise and disentangled control over multiple extra facial motions, which can hardly be achieved by previous methods.

cs.CV↗

arXiv2022

Phonemic Adversarial Attack against Audio Recognition in Real World

Recently, adversarial attacks for audio recognition have attracted much attention. However, most of the existing studies mainly rely on the coarse-grain audio features at the instance level to generate adversarial noises, which leads to expensive generation time costs and weak universal attacking ability. Motivated by the observations that all audio speech consists of fundamental phonemes, this paper proposes a phonemic adversarial tack (PAT) paradigm, which attacks the fine-grain audio features at the phoneme level commonly shared across audio instances, to generate phonemic adversarial noises, enjoying the more general attacking ability with fast generation speed. Specifically, for accelerating the generation, a phoneme density balanced sampling strategy is introduced to sample quantity less but phonemic features abundant audio instances as the training data via estimating the phoneme density, which substantially alleviates the heavy dependency on the large training dataset. Moreover, for promoting universal attacking ability, the phonemic noise is optimized in an asynchronous way with a sliding window, which enhances the phoneme diversity and thus well captures the critical fundamental phonemic patterns. By conducting extensive experiments, we comprehensively investigate the proposed PAT framework and demonstrate that it outperforms the SOTA baselines by large margins (i.e., at least 11X speed up and 78% attacking ability improvement).

cs.SD↗

arXiv2022

Towards Comprehensive Testing on the Robustness of Cooperative Multi-agent Reinforcement Learning

While deep neural networks (DNNs) have strengthened the performance of cooperative multi-agent reinforcement learning (c-MARL), the agent policy can be easily perturbed by adversarial examples. Considering the safety critical applications of c-MARL, such as traffic management, power management and unmanned aerial vehicle control, it is crucial to test the robustness of c-MARL algorithm before it was deployed in reality. Existing adversarial attacks for MARL could be used for testing, but is limited to one robustness aspects (e.g., reward, state, action), while c-MARL model could be attacked from any aspect. To overcome the challenge, we propose MARLSafe, the first robustness testing framework for c-MARL algorithms. First, motivated by Markov Decision Process (MDP), MARLSafe consider the robustness of c-MARL algorithms comprehensively from three aspects, namely state robustness, action robustness and reward robustness. Any c-MARL algorithm must simultaneously satisfy these robustness aspects to be considered secure. Second, due to the scarceness of c-MARL attack, we propose c-MARL attacks as robustness testing algorithms from multiple aspects. Experiments on \textit{SMAC} environment reveals that many state-of-the-art c-MARL algorithms are of low robustness in all aspect, pointing out the urgent need to test and enhance robustness of c-MARL algorithms.

cs.MA↗

arXiv2022

Defensive Patches for Robust Recognition in the Physical World

To operate in real-world high-stakes environments, deep learning systems have to endure noises that have been continuously thwarting their robustness. Data-end defense, which improves robustness by operations on input data instead of modifying models, has attracted intensive attention due to its feasibility in practice. However, previous data-end defenses show low generalization against diverse noises and weak transferability across multiple models. Motivated by the fact that robust recognition depends on both local and global features, we propose a defensive patch generation framework to address these problems by helping models better exploit these features. For the generalization against diverse noises, we inject class-specific identifiable patterns into a confined local patch prior, so that defensive patches could preserve more recognizable features towards specific classes, leading models for better recognition under noises. For the transferability across multiple models, we guide the defensive patches to capture more global feature correlations within a class, so that they could activate model-shared global perceptions and transfer better among models. Our defensive patches show great potentials to improve application robustness in practice by simply sticking them around target objects. Extensive experiments show that we outperform others by large margins (improve 20+\% accuracy for both adversarial and corruption robustness on average in the digital and physical world). Our codes are available at https://github.com/nlsde-safety-team/DefensivePatch

cs.CV↗

arXiv2021

Dual Attention Suppression Attack: Generate Adversarial Camouflage in Physical World

Deep learning models are vulnerable to adversarial examples. As a more threatening type for practical deep learning systems, physical adversarial examples have received extensive research attention in recent years. However, without exploiting the intrinsic characteristics such as model-agnostic and human-specific patterns, existing works generate weak adversarial perturbations in the physical world, which fall short of attacking across different models and show visually suspicious appearance. Motivated by the viewpoint that attention reflects the intrinsic characteristics of the recognition process, this paper proposes the Dual Attention Suppression (DAS) attack to generate visually-natural physical adversarial camouflages with strong transferability by suppressing both model and human attention. As for attacking, we generate transferable adversarial camouflages by distracting the model-shared similar attention patterns from the target to non-target regions. Meanwhile, based on the fact that human visual attention always focuses on salient items (e.g., suspicious distortions), we evade the human-specific bottom-up attention to generate visually-natural camouflages which are correlated to the scenario context. We conduct extensive experiments in both the digital and physical world for classification and detection tasks on up-to-date models (e.g., Yolo-V5) and significantly demonstrate that our method outperforms state-of-the-art methods.

cs.CV↗